DTIC ADP012596: Raman and Photoluminescence Mapping of Lattice Matched InGaP/GaAs Heterostructures

7 Pages·0.38 MB·English

by  Defense Technical Information Center

UNCLASSIFIED
Defense Technical Information Center
Compilation Part Notice
ADP012596
TITLE: Raman and Photoluminescence Mapping of Lattice Matched
InGaP/GaAs Heterostructures
DISTRIBUTION: Approved for public release, distribution unlimited
This paper is part of the following report:
TITLE: Progress in Semiconductor Materials for Optoelectronic
Applications Symposium held in Boston, Massachusetts on November
26-29, 2001.
To order the complete compilation report, use: ADA405047
The component part is provided here to allow users access to individually authored sections
)f proceedings, annals, symposia, etc. However, the component should be considered within
[he context of the overall compilation report and not as a stand-alone technical report.
The following component part numbers comprise the compilation report:
ADP012585 thru ADP012685
UNCLASSIFIED
Mat. Res. Soc. Symp. Proc. Vol. 692 © 2002 Materials Research Society H2.11
RAMAN AND PHOTOLUMINESCENCE MAPPING OF LATTICE MATCHED
InGaP/GaAs HETEROSTRUCTURES
G. Attolini, P.Fallini, F.Germini, C. Pelosi
MASPEC-CNR Institute, Parco Area Delle Scienze, 37/A Fontanini, 43010 Parma, Italy
0. Martfnez, L.F.Sanz, M.A. Gonzdlez, J. Jim6nez
Ffsica de ]a Materia Condensada, ETSII, 47011 Valladolid, Spain
ABSTRACT
The influence of the substrate on composition and CuPt-type spontaneous order of
MOVPE lattice matched InGaP/GaAs layers was studied. The study was carried out by
microRaman and microphotoluminescence. The order was determined by the band gap, while the
Raman parameters were also contributed by the surface topography that was also related to the
type of substrate. The spontaneous order increases with Si- doping of the substrates. Doping the
layers with Zn randomises the alloy.
INTRODUCTION
InGaP layers lattice matched to GaAs substrates have potential applications for electronic
devices, e.g. high efficiency tandem solar cells, single heterojunction bipolar transistors, tunable
laser diodes, etc. It has some advantages in relation to AlGaAs ternary alloys, which are the base
of many devices. Lattice matched InGaP presents a direct band gap of 1.9 eV, which is
approximately equal to the maximum direct band gap that can be obtained with AIGaAs (Ga
molar fraction of 0.4) [1]. InGaP is insensitive to oxygen and humidity inside the reactor, which
is a well known cause of Al instability in AIGaAs [2].
InGaP is affected by other problems, in particular the composition control and the
existence of an spontaneous ordered phase. In xGaxP matches the GaAs lattice for x=0.516 at
1
room temperature; however due to the large lattice and thermal mismatches between InP, GaP
and GaAs the composition is critical to avoid residual strain in the layers [3]. The second
problem of interest regards the spontaneous order. InGaP can appear under a CuPt-type ordered
phase [4, 5]. In the ordered phase the cations are not randomly distributed, but seggregate
spontaneously into alternate (111) planes giving a CuPt-type structure in the cation sublattice.
This phase critically influences the optical and electrical properties of InGaP/GaAs layers. The
most important is the shrinking of the band gap, which can be reduced by 100 meV for lattice
matched composition [5, 6]. Also, the minority carrier lifetime and mobility are sensitive to
cation order [6]. The spontaneous order is related to the growth method and the specific growth
conditions, such as the substrate temperature and the V/III ratio. The nature of the susbstrate is
shown here to influence the properties of the layers.
We present herein a detailed study of the properties of InGaP layers grown by LP-MOVPE
(Low pressure metal organic vapor phase epitaxy) on different substrates aiming to understand
the influence of the substrate on the spontaneous order. The analysis is carried out by High
Resolution X-Ray diffraction (HRXRD), Atomic Force Microscopy (AFM), micro-Raman (gt-R)
and micro-Photoluminescence (g-PL).
91
EXPERIMENTAL AND SAMPLES
Layers were grown at reduced pressure (60 mBar) by MOVPE in a horizontal reactor. The
precursors were trimethylgallium (TMG), trimethylindium (TMI) and Phosphine (PH3) and
Arsine (AsH ) as main reagents; arsine was diluted in hydrogen (5%). The layers of InGaP were
3
deposited on (100) GaAs substrates cutted 2' off towards the [110] axis. The growth conditions
were the same for all runs, namely substrate temperature of 600'C and V/Ill ratio of 159.60
(rel.u.). The substrates were either semiinsulating (SI) or n-type (Si-doped), the doping level of
the substrate was monitored by Raman spectroscopy, using the LO phonon-plasmon coupled
(LOPC) modes; samples B, D, I and J were grown on semiinsulating substrates, samples F and H
were grown on Si-doped substrates ([Si]>3xl015 cm'). The InGaP layers were either undoped
(B, D, F and H) Si (J) or Zn (I) doped.
The AFM images show that the surface morphology of the InGaP layers was determined
by the type of substrate, figure 1. In particular, layers grown on n-type substrates present flat
surfaces suggesting layer by layer growth. Instead of this, layers grown on SI substrates
exhibited a rough surface, looking like an array of parallel ridges about 0.2 pm wide, which
exposes (100), (101) and (Il l) planes.
a
Figure 1. AFM images of samples grown on n-type substrates (Si-doped) (a), and semiinsulating
substrates (b)
The layers were investigated by HRXRD. Two (R=0,t) symmetric 004 reflections, R being
the azimuthal angle, have been measured from each sample to eliminate the effect of a possible
miscut angle. All the diffraction profiles were recorded in the w.20 scan mode. To estimate the
perpendicular mismatch (Adid)' we used the first order formula (Adld)' = -A(cootgB , where OB
is the Bragg angle of the substrate and Awthe peak separation in arcsec.
p-R measurements were carried out with a DILOR XY Raman spectrometer attached to a
metallographic microscope. The 514.9 nm line of an Ar÷ laser was focused onto the sample by
the large numerical aperture (NA=0.95) of the microscope objective, which also collected the
scattered light, conforming a nearly backscattering geometry. In our usual experimental
conditions the laser beam diameter at the focal plane was slightly sub-micrometric.
Room temperature luminescence spectra were obtained in the same DILOR XY Raman
spectrometer, thus warranting that the same points of the samples were measured in Raman and
Luminescence.
92
RESULTS AND DISCUSSION
The composition of the layers was estimated from HRXRD and g-R data. The
perpendicular lattice mismatches, (Ad/ld)L, estimated for the six samples studied were positive,
which corresponds to compressive mismatch and therefore to slightly Ga rich stoichiometry, the
composition was close to the lattice matched value since the maximum value of (Ad/ld)' was
5. 10-.
g-R spectra were obtained on different points of the samples. The g-PL spectra were
obtained on the same points, which allowed a correlation between both measurements. The
Raman spectrum of InGaP presents a two mode behaviour [7]. It consists of an LO phonon mode
(GaP like) at 380 cm-1 (henceforth labeled L0 ), a TO phonon mode (InP-like, TO ) at 330 cm-l,
1 2
which is selection rule forbidden, but is activated by alloy disorder, figure 2. Special mention
should be paid to the phonon band at 365 cm', which has been associated with the InP-like LO
phonon (LO ), the properties of this band have been reported to be sensitive to the spontaneous
2
order. This band seems to have a complex character, in fact a TO band, GaP-like (TO ) seems to
1
be located in the same spectral region [8]. On the other hand, the frequencies and relative
intensities of the modes are dependent on the composition.
There is a clear difference between the Raman spectra of doped and undoped layers. The
phonon bands are broadened and the forbidden TO band is enhanced in the doped layers. This is
the consequence of the existence of LOPC modes [8]. Therefore, a comparison between the
Raman spectra of doped and undoped layers must be done with caution.
The LO, phonon can be used to determine the composition of the layers since its Raman
shift is weakly affected by the spontaneous order [7]. The LO, phonon frequency shifts almost
linearly with x; it increases by 0.7 cm'/%Ga, in the lattice matched composition range [9].
According to the X-Ray data the samples were slightly Ga-rich. This is confirmed by Raman
spectroscopy, since the Raman frequencies were all above the Raman frequency for lattice
matching, which is 382 cm'; therefore the layers were slightly Ga-rich (- 52% Ga). The average
Raman shifts measured are 382.83 cm-(sample B), 382.85 cm' (sample D), 382.52 cm-1 (sample
F), 382.37 cm-1 (sample H), 382.50 cm' (sample I) and 385.89 cm"1 (sample J). Samples B, D, F,
H and I have compositions that differ each other by less than 1%. Sample J presents a large
Raman shift that can be interpreted as a Ga rich composition (z55%).
The Raman scattering is sensitive to the atomic arrangement; in fact, spontaneous CuPt-
type order was observed to influence the Raman spectrum. An empirical relation between the
band gap and the depth of the valley between the two LO peaks in the Raman spectrum was
reported [10]. The valley depth normalized to the intensity of the L0 phonon peak is taken as an
2
order parameter, labeled S, see figure 2. The existence of a profound valley between the two LO
bands (high S) corresponds to a large band gap (disordered layer), a value of 0.6 has been
reported for fully disordered specimens. The valley depth decreases for partially ordered
samples, reaching a value of 0.2 for fully ordered samples [7,10].
The Raman spectra were obtained at several points of the layers using the Raman
microprobe. The Raman parameters of the spectra, that is, Raman Shifts OLOj and OfLO2 and the
FWHM (Full Width at Half Maximum) of the modes were studied in relation to the valley depth
aiming to establish a correlation between spontaneous order and the Raman parameters. The
FWHM of the peaks is related to the phonon correlation length. In fact, the FWHM should
decrease when In and Ga are randomly distributed in the cation sublattice and will increase for
93
clustering or in the presence of the ordered phase. The FWHM of the L0 band was observed to
1
increase with the valley depth, which is controversial with the above interpretation. A good
correlation is found for the frequency splitting between both LO modes, AOQ.o=(0,ol-oO ). and
2
the valley depth, figure 3. The valley depth increases when this splitting increases. This cannot
be related to composition changes since the major contribution to this splitting is due to the down
frequency shift of the L0 Raman band, the L0 band frequency remaining within the limits
2 1
determined by the small composition variation above discussed, figure 3. According to the above
discussion, a shift to the low frequency of OWo2 takes place with a broadening of the L0 band.
1
These data are only referred to the undoped samples, since the spectra obtained for doped
samples, I and J, are influenced by the LOPC modes.
--250 LO 3 375
5 370 .
2100- 305
D 050
L% E
0 25 365
TO,
50 DALA 300360
b 20
250 30D 400 150 .10 0.15 0.20 0.25 0.30 0.35 0.43055
Raman shift (cm"1) S (valley depth to peak ratio)
Figure 2.. Typical Raman spectra of lattice Figure 3.. A(o.o((.ot-owo2) (e) and
matched InGaP/GaAs. The order parameter, S, ",02 (lnP-like) (0) vs S
is defined as the valley depth, d, normalized to
the intensity of the L0 band, b ( S=d/b)
2
Taking the band gap as the peak energy of the intrinsic PL band at room temperature one
obtains the following band gap average energies for the samples: 1.90 eV (sample B), 1.87 eV
(sample D), 1.85 eV (sample F), 1.86 eV (sample H), 1.91 eV (sample I) and 1.89 eV (sample J).
The PL spectra were measured at the same points that the Raman spectra.
The band gap is also influenced by the layer composition. In the lattice matched region the
band gap is enlarged by 15 meV for an increase of 0.01 in the molar fraction of Ga [II].
However, the composition of samples B, D, F, H and I was observed to lay within less than 1%
percent interval. Therefore, one can assume that the band gap differences between these samples
are mostly due to the degree of spontaneous order. Only the band gap measured for sample J
could be additionally contributed by composition deviation from the lattice matched value,
according to the Raman data. The largest band gap and therefore the lowest degree of
spontaneous order corresponds to samples B and I, followed by sample D and finally the higher
degree of order corresponds to samples F and H. Taking into account the characteristics of the
samples one can argue that: i) n-type substrates benefit the spontaneous order, the degree of
order increases with the dopant concentration of the substrate, ii) SI substrates reduce the degree
of spontaneous order, iii) Zn doping randomizes the alloy, which is a well known property of Zn
[12].
If one plots the band gap vs the order parameter, S, for the undoped samples one observes
that the expected correlation does not appear, figure 4. The samples with the largest band gap, B
94
and D, have small S, while the samples with the smallest band gap have higher values of S. This
result is controversial with the results reported by other authors. This can be explained on the
basis of the complex nature of the Raman band labeled L0 and the surface topography of our
2
samples. The samples grown on SI substrates present a rough topography with high index planes,
while the topography of the samples grown on n-type substrates was smooth, see figure 1. In our
scattering geometry the TO bands are forbidden, however, they can be activated by either alloy
fluctuation or the surface orientation. The presence of high index planes in samples B and D
should activate the forbidden TO, mode, with the corresponding lineshape change of the band at
365 cm', up to now called L0 , which would result on a reduction of S in spite of the lower
2
spontaneous CuPt-type order of these samples. The smooth surface of samples F and H should
reduce the contribution of the forbidden TO, mode to the L0 Raman band. An attentive
2
observation of the spectra of samples B and D shows that the decrease of S is accompanied by an
increase of the intensity of the other TO phonon band (InP-like, TO ) at 330 cm-1, figure 5,
2
which is consistent with the activation of the TO forbidden modes and the contribution of the
TO mode to the value of S.
1
0.40 0.40
0.40 .3
0.3 v* S OBD 0.5vC C v DF
0.30 F 0.30 H
O H
(0) 0.25 00 (00.25
0.20 0.20
0.15 0.15
0.10 0.10
1.85 1.86 1.87 1.88 1.89 1.90 1.91 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
E,(eV) IT0/It0
Figure 4. S vs Eg(nm). The line is a guide Figure 5. S VS IT02 /IL02, TO is the band at
2
to the eye 330cm'. The straight line is a guide to the eye.
Finally, monochromatic (intrinsic emission) intensity PL maps were obtained for the
different samples at room temperature. Intensity fluctuations were observed, figure 6. The
intrinsic PL intensity is governed by the competition between band to band recombination and
non radiative recombination at deep levels. Also a decrease of the luminescence intensity has
been reported (up to a factor 3) for the ordered phase [11]. Raman data did not reveal significant
composition changes over a sample. However, a correlation between the PL intensity and the
peak energy was observed in samples grown on n-type substrates, see Fig. 6. Therefore, the non-
uniformities of the PL intensity and peak wavelength can be understood if one considers that the
PL intensity and band gap energy changes are controlled by the ordered/disordered volume ratio
probed by the laser beam.
The influence of the substrate on the characteristics of the samples can be tentatively
explained by the change in the lattice parameter introduced by the Si impurities, since the other
growth parameters were the same for all the growth runs. The smooth layer by layer growth
allows the segregation of cations, while the tridimensional growth avoids it.
95
iI E 0.7
'(cid:127) .. . . :: a ob
I.N8 1.81 12 .3 14 1W
Energy (.V)
Figure 6. Monochromatic PL map (room temperature) of sample F, (a). The bar is 1 mm. IPL vs
Eg measured in sample F (b).
CONCLUSIONS
The influence of the type of substrate on the order of lattice matched lnGaP/GaAs layers
was studied by p-R and p-PL. The Si doping of the substrates benefits the ordering, which is
related to the shrink of the band gap. The Raman parameters were demonstrated to be also
influenced by the surface topography, since the S parameter was related to the activation of the
forbidden TO (GaP-like) Raman mode, therefore, S cannot be used as a measure of the order for
1
samples with rough surface. Spatial inhomogeneities were related to the distribution of the
ordered / disordered domains.
ACKNOWLEDGEMENTS. The Spanish group was funded by JCL (Junta de Castilla y Le6n) and FES (Fondo
Social Europeo) (Project: VA047/01). The Italian group acknowledges the financial support of the Italian Space
Agency.
REFERENCES
1. J.R. Shealy, C.F. Schaus, L.F. Eastman; Appl. Phys. Lett. 48, 242 (1986)
2. T.F. Kuech. D.J. Wolford, E. Venhoff, V. Deline, P.M. Mooney, R. Potemski, J. Bradley;
J.Appl. Phys. 62, 632 (1987)
3. Y.Q. Wang, Z.L. Wang, T. Brown, A. Brown, G. May; J. Electron. Mater. 29, 1372
(2000)
4. Krost, N. Esser, H. Selber, J. Christen, W. Richter, D. Bimberg, L.C. Su, G.B.
Stringfellow; J. Vac. Sci. Technol. 12, 2558 (1994)
5. S.F.Yoon. K.W.Mah, H.Q.Zheng; Opt. Mater.14, 59 (2000)
6. Sasaki, K. Tsuchida, Y. Narukawa, Y. Kawakami, S.G. Fujita. Y. Hsu, G.B. Stringfellow;
J.Appl. Phys. 89, 343 (2001)
7. F.Alsina, N.Mestres, J.Pascual, C.Geng, P.Ernst, F.Scholz; Phys.Rev. B 53, 12994 (1996)
8. K.Sinha, A. Mascarenhas, S.R.Kurtz, J.M.Olson; J.Appl.Phys. 78, 2515 (1995)
9. M. Zachau, W.T. Masselink; Appl. Phys. Lett. 60, 2098 (1992)
10. T. Suzuki. A. Gomyo, S. Iijima, K. Kobayhasi, S. Kawata, T. Hino, T. Yuasa; Jpn. J.
Appl. Phys. 27, 2098 (1988)
11. G.H. Olsen, C.J. Nuese, R.T. Smith; J. Appl. Phys. 49, 5523 (1978)
12. T. Suzuki, A.Gomyo, K.Kobayhasi, S. Kawata, S.Iijima; .Jpn. .1. Appl. Phys. 27, L1549
(1988)
96